A Hybrid ANN-SNN Architecture for Low-Power and Low-Latency Visual Perception

arXiv:2303.14176v2 Announce Type: replace
Abstract: Spiking Neural Networks (SNN) are a class of bio-inspired neural networks that promise to bring low-power and low-latency inference to edge devices through asynchronous and sparse processing. However, being temporal models, SNNs depend heavily on expressive states to generate predictions on par with classical artificial neural networks (ANNs). These states converge only after long transient periods, and quickly decay without input data, leading to higher latency, power consumption, and lower accuracy. This work addresses this …

abstract ann architecture artificial arxiv asynchronous bio bio-inspired class cs.ai cs.cv devices edge edge devices generate however hybrid inference latency low networks neural networks perception power predictions processing snn spiking neural networks temporal through type visual